| Summary: | 碩士 === 國立中興大學 === 土木工程學系 === 92 === Due to the progressive development in compute science and image processing technology in recent years, the automatic detection of road network from digital images has become successfully feasible. Aerial images and remote sensing images are primary sources for use in such image-based road network extraction techniques.
The purpose of this study is to establish a new model for road network detection based on digital image processing techniques. In this study, large scale aerial images are used to detect the road marks in order to build corresponding road network. Road marks follow the trend of roads. If the road marks could be extracted from aerial images, the frame of the global road network can be established from them. Road marks have specific spectral characteristics and regularity. The scheme of road network extraction in this research involves supervised classification to extract road mark regions, road segmentation and noise removal by region growing and masking, and least squares curve fitting to establish the trend of road network.
The results of experiments show that the proposed method in this research can successfully extract road marks from large scale aerial images and obtain the conditions for road network establishment. These results are contributive to the development of robustic algorithms for automatic extraction of accurate road networks from aerial images.
|
|---|
